De Novo Missense Variants in TRAF7 Cause Developmental Delay, Congenital Anomalies, and Dysmorphic Features

Abstract

Main Text

Somatic and germline mutations in proto-oncogenes and tumor suppressor genes are well-known causes of cancer. Many proto-oncogenes and tumor suppressor genes also play a crucial role in human development and as such, germline mutations in these genes can lead to developmental disorders.^{1, 2, 3, 4, 5, 6} TRAF7 (Tumor necrosis factor receptor-associated factor 7 [MIM: 606692]) belongs to the multi-functional TRAF family and is involved in multiple biological processes, including ubiquitination and myogenesis.⁷ TRAF7 is also a known mediator of the MAP kinase and NF-κB signaling pathways.⁷ Somatic mutations in TRAF7 have been reported in meningioma and mesothelioma.⁸ Although germline TRAF7 variants have been reported in two individuals with autism spectrum disorder,^{9, 10} a clear association between germline defects in TRAF7 and human disease has not been established.

Herein we report missense variants in TRAF7 in seven unrelated individuals with overlapping features. This study was performed according to the standards of the ethics committees and the institutional review boards at Baylor College of Medicine, the National Institutes of Health, and Stanford University. Written informed consent was obtained from all study participants. Medical records were reviewed and clinical information was extracted for phenotype analysis. Key clinical features are summarized in Table 1 and are discussed below; comprehensive phenotype information is available in Table S1. Select photographs of the described subjects are presented in Figure 1. Clinical exome sequencing was performed as previously described;¹¹ sequencing metrics are summarized in Table S2.

Open in a separate window

Figure 1

Facial Features and Limb Phenotypes for Subjects with TRAF7 Variants

Shown are subject 7 (A), subject 4 (B), subject 3 (C), subject 5 (D), subject 6 (E), and subject 1 (F). Shared dysmorphic features included epicanthal folds, ptosis, abnormally set or dysplastic ears, a low hairline, excess nuchal skin, and multiple hair whorls. Overlapping toes are shown in (C) and (F).

The study cohort consisted of two female and five male subjects ranging in age from 1 week to 43 years. Prenatal histories were notable for antenatal detection of heart anomalies (n = 3), cystic hygroma (n = 2), and two-vessel cord (n = 2). Postnatally, several patients had respiratory distress (n = 4) and poor feeding (n = 3). Congenital heart defects were present in six of seven patients and ranged in type and severity (Tables 1 and S1).

Motor and/or speech delay were present to a variable degree in five of the five subjects for whom this outcome could be assessed. Subjects 3 and 7 had documented seizures, and subject 6 had a history of staring spells concerning for absence seizures. Brain imaging revealed variant venous anatomy in two subjects and prominent/enlarged ventricles and/or cerebral atrophy in four subjects (Table S1). Subject 6 was reported to have sulcal/gyral asymmetry of the right frontal parietal region with mild cortical irregularity concerning for possible polymicrogyria. A tethered cord was found in subjects 1, 3, and 4. Sensorineural and mixed hearing loss were reported in subjects 1 and 3, respectively. Cortical blindness was reported in subject 2 and optic atrophy was reported in subjects 2 and 3.

Shared dysmorphic features included epicanthal folds (n = 5), ptosis (n = 4), abnormally set or dysplastic ears (n = 6), and widely spaced and/or inverted nipples (n = 5) (Figure 1). A low hairline and/or excess nuchal skin were reported in five subjects, two of whom had a prenatal history of a cystic hygroma. Three subjects were found to have multiple hair whorls.

Limb and digital anomalies were present in all seven subjects to a variable degree, including variant palmar/digital crease patterns (n = 6) and overlapping toes (n = 5). Subject 3, the eldest of the cohort, had relatively extensive involvement of the musculoskeletal system including a history of scoliosis with degenerative joint disease, an asymmetric sternum, bilateral avascular necrosis of the hip, tibial malformation, valgus deformity of the ankles, flexion contractures of the hands (Figure 1C), and subluxation of multiple joints in the hands and feet.

In summary, the seven individuals described herein share similar clinical features including developmental delay (5/5), congenital heart defects (6/7), limb and digital anomalies (7/7), and dysmorphic facial features (7/7).

TRAF7 is a 21-exon gene located at chromosomal region 16p13.3. The longest isoform of the TRAF7 protein (GenBank: {"type":"entrez-protein","attrs":{"text":"NP_115647","term_id":"45594312","term_text":"NP_115647"}}NP_115647) is 670 amino acids in length and contains an N-terminal ring finger domain, an adjacent zinc-finger domain, a centrally situated coiled-coil motif, and seven WD40 repeats in the C-terminal domain (Figures 2B and 2C).^{7, 12} Clinical exome sequencing identified a total of four distinct missense variants in TRAF7 (GenBank: {"type":"entrez-nucleotide","attrs":{"text":"NM_032271.2","term_id":"45594311","term_text":"NM_032271.2"}}NM_032271.2) in the seven described subjects, including a recurrent c.1964G>A (p.Arg655Gln) variant found in four subjects. All available parental samples were analyzed by exome and/or Sanger sequencing to assess inheritance of the TRAF7 variants. The variants occurred de novo in six subjects (subjects 1–3 and 5–7). For subject 4, a maternal sample was negative for the variant but a paternal sample was not available for testing.

Open in a separate window

Figure 2

In Silico Analysis of the Four TRAF7 Variants Identified in This Study

(A) The R655, T601, R371, and K346 residues are conserved from human to zebrafish (prepared from UCSC genome browser Multiz Alignments of 100 Vertebrates track).

(B) Schematic view of the TRAF7 exon-intron structure. Blue boxes represent exons and yellow fields represent introns. The identified cDNA changes are listed.

(C) Schematic view of the TRAF7 protein and its domains based on data extracted from Zotti et al.⁷ Domains are represented by blue shapes. The domain names and the identified amino acid changes are listed.

The four TRAF7 variants detected in our cohort are absent from the ExAC and gnomAD databases (access date 5/1/2018) and alter highly phylogenetically conserved amino acids (Figure 2A). The in silico variant effect prediction tool MutationTaster predicts all four variants to be disease-causing,¹³ while SIFT and PolyPhen-2 yield consistent results with two exceptions (Table 2).^{14, 15} Subjects 1–4 had a c.1964G>A (p.Arg655Gln) variant in exon 20 that localized to the seventh of seven WD40 domains at the C terminus of TRAF7 (Figures 2B and 2C). The c.1964G>A (p.Arg655Gln) variant has been previously reported as a de novo event in a female with autism spectrum disorder.⁹ The recurrent nature of this variant is likely explained by its location within a CpG dinucleotide. Subject 5 had a c.1801A>G (p.Thr601Ala) variant which localized to exon 19 and the sixth WD40 domain. Subject 6 had a c.1036A>G (p.Lys346Glu) variant found in exon 11 affecting an amino acid in the coiled-coiled motif. Finally, subject 7 had a mosaic c.1111C>G (p.Arg371Gly) variant in exon 12 that localized to the coiled-coiled domain. The c.1111C>G (p.Arg371Gly) variant had an allele fraction (AF) of 13% (14 mutant and 105 total reads) by exome sequencing using DNA extracted from the blood of this individual. As a confirmatory study, we used the restriction enzyme, HhaI, to digest the reference allele in both proband and parental samples then Sanger sequenced the digested product. This analysis demonstrated unequivocal presence of the c.1111C>G allele in the proband but not the parental samples, confirming the presence of the mosaic variant in the proband only (Figure S1). We further performed PCR followed by high-throughput sequencing using DNA extracted from the blood of this individual and the parents (Supplemental Material and Methods). We found that the c.1111C>G variant had an AF of 15% at 23,108× coverage in the proband and an AF of 0% at 22,523× and 22,957× coverage in the mother and father, respectively. Notably, the phenotype of subject 7 is not apparently milder than the other six individuals. This may be explained by variability in the levels of mosaicism in critical tissues.

Using different deletion mutations, several groups have clarified the functional relevance of the recognizable TRAF7 protein domains. The C-terminal WD40 domain has been shown to facilitate TRAF7 interaction with MAP3K3 (MIM: 602539) while TRAF7 homodimerization and subcellular localization appear to rely on the presence of the centrally located coiled-coiled and zinc finger motifs.^{12, 16} Of the four TRAF7 mutations detected in our cohort, two are found within the coiled-coiled domain and two in the WD40 repeat region (Figure 2C). Additionally, the p.Lys346Glu variant detected in subject 6 affects an experimentally determined ubiquitination site.¹⁷ Finally, TRAF7 is predicted to be intolerant to missense variations (z = 3.34).¹⁸ Thus, the localization of the detected TRAF7 variants to highly conserved amino acids in key functional domains and the predicted intolerance of TRAF7 to missense variation support the notion that missense mutations in TRAF7 may be causative of a human phenotype.

We sought to evaluate the likelihood of detecting de novo variants in TRAF7 and the recurrent de novo c.1964G>A change in our exome database by chance. To do so, we performed a gene-specific mutation rate analysis using a previously published approach.¹⁹ In total there were about 7,700 individuals in our undiagnosed exome cohort at the time of analysis. Focusing on rare variants (population frequency < 1% in gnomAD database) in this cohort, we identified seven individuals with de novo variants in TRAF7 (including subjects 1–3, 5–7, and one additional individual for whom we did not obtain research consent) and three individuals with the recurrent de novo c.1964G>A (p.Arg655Gln) variant (including subjects 1–3 and excluding subject 4 for whom the paternal sample was not available). We used 2.4 × 10⁻⁵ as the expected TRAF7 gene-specific rate of de novo missense mutation, which was previously determined based on gene length, local sequence context, and mutation type.²⁰ After adjusting the p value for 18,892 genes in the Consensus Coding Sequence for genome-wide statistical significance, the likelihood to detect seven individuals with de novo TRAF7 variants and three individuals with the recurrent de novo c.1964G>A (p.Arg655Gln) variant in our cohort was 2.6 × 10⁻³ and 1.9 × 10⁻⁸, respectively (Table S3). These data indicate that our undiagnosed exome cohort is enriched for de novo variants in TRAF7 and renders improbable the detection of seven de novo TRAF7 variants in a cohort this size by chance.

Seven TRAF (TNF receptor associated factor) proteins (TRAF1–7) have been characterized in mammals to date and comprise the TRAF family of proteins.⁷ TRAF proteins are intracellular adaptors that bind TNF family receptors and transduce the cellular effects mediated by TNF family ligands.^{7, 21} Mitogen-activated protein kinases (MAPKs) comprise a family of protein kinases that regulate several cellular activities including gene expression, mitosis, movement, metabolism, and programmed death.²² There are three subfamilies of MAPKs in multicellular organisms: ERK1 and ERK2; c-Jun NH2-terminal kinases JNK; and p38.²² Although the phenotypic effects of variants in TRAF7 have not been clearly characterized in humans or in model organisms to our knowledge, a role for TRAF7 in MAPKs pathways has been clearly established. Bouwmeester et al. and Xu et al. showed that TRAF7 interacts by its WD40 domain with MAP3K3 and that coexpression of TRAF7 with MAP3K3 activates the JNK and p38 signaling pathways that affect cell proliferation, differentiation, survival, and migration.^{12, 16, 23} The ERK1 and ERK2 signaling cascades have not been directly linked to TRAF7, but both are activated by MAP2K1/MAP2K2 (MEK1/2)²⁴ which in turn are activated by multiple protein kinases including MAP3K3^{25, 26}—a known TRAF7 binding partner. Collectively, these lines of evidence suggest that the TRAF7 variants detected in our cohort may interfere with TRAF7 regulation of MAPK pathways.

To characterize the functional consequence of the TRAF7 variants detected in the described subjects, we performed site-directed mutagenesis to introduce the four observed variants into TRAF7 cDNA, which we then transfected into HEK293T cells (Supplemental Material and Methods). Expression of mutant TRAF7 had no apparent effect on total or phosphorylated p38 and JNK when assessed by immunoblotting (data not shown). Cellular levels of ERK1/2, mediators of a third MAPK cascade, were not affected by mutant TRAF7, but phosphorylation of ERK1/2 was reduced in mutant-transfected cells relative to wild-type (Figure 3A). This reduction was statistically significant with TNFα stimulation for cells expressing the p.Arg655Gln, p.Thr601Ala, or p.Arg371Gly variants (Figures 3D and 3E). Cells expressing the p.Lys346Glu variant demonstrated a consistent but not statistically significant reduction in phosphorylated ERK1/2 (Figures 3B–3E).

Open in a separate window

Figure 3

The Effect of TRAF7 Mutants on ERK1/2 Signaling

(A) Phosphorylation of ERK1/2 was analyzed by transiently expressing the FLAG-tagged wild-type and mutant constructs in HEK293T cells. Twenty-four hours after transfection, cells were stimulated with 10 ng ml⁻¹ TNFα for 30 min. Cells were then collected and lysed and immunoblotting was performed. Expression of the TRAF7 mutants resulted in decreased phosphorylation of ERK1/2 when compared with cells expressing wild-type protein, either with or without TNFα treatment. Asterisk and pound symbols indicate ERK1 and ERK2, respectively.

(B–E) Quantification of the effect of TRAF7 mutants on ERK1/2 signaling. pERK1 (B) and pERK2 (C) signaling without TNFα treatment. pERK1 (D) and pERK2 (E) signaling with 10 ng ml⁻¹ TNFα stimulation for 30 min.

All data were normalized to GAPDH protein levels, with the wild-type protein set at 1.0. The results are representative of three independent experiments. ^∗p ≤ 0.05, ^∗∗p ≤ 0.01, ^∗∗∗p ≤ 0.001, ^∗∗∗∗p ≤ 0.0001, one-way ANOVA with Dunnett’s multiple-comparisons test. Data were shown as mean ± standard error of the mean, n = 3.

ERK1/2, also known as MAPK3/MAPK1, act downstream of Ras²⁴ and mobilize a signaling cascade involved in cell proliferation, differentiation, and survival.²⁷ Functional redundancy between ERK1 and ERK2 has been demonstrated in mice, but ERK1 knockout mice are viable while ERK2 knockout mice die by E11.5 due to defective placental development.²⁸ Conditional knockout of ERK2 in the embryo while sparing the placenta results in embryonic survival to E16.5.²⁷ Interestingly, ERK2 conditional knockout mice show a phenotype highly consistent with what we have observed in our subjects with TRAF7 variants, including craniofacial abnormalities, cardiovascular malformations, and limb defects.²⁷ Additionally, conditional knockout or knockdown of ERK2 in the mouse central nervous system leads to reduced cortical thickness, profound deficits in associative learning, abnormal behaviors related to autism spectrum disorder and anxiety, and impaired long-term memory.^{29, 30, 31} Thus, the mouse models support that reduced ERK1/2 activation resulting from mutations in TRAF7 represents a plausible mechanism of disease in the described subjects. Further study is needed to illustrate the mechanism of TRAF7 regulation on ERK1/2 and MAPKs pathways.

Interestingly, subject 1 had been evaluated for a possible diagnosis of cardio-facio-cutaneous syndrome and subject 2 was suspected of having a Noonan-spectrum condition prior to the detection of the TRAF7 variants by exome sequencing (Table S1). On further review, we noted that subject 3 has features reminiscent of Costello syndrome including a coarse facial appearance, congenital heart defect, and musculoskeletal involvement including scoliosis and contractures. Further studies are needed to illustrate the overlapping and discordant phenotypic features and the possible mechanistic link between RASopathy and TRAF7-related disorder. Our results, along with the studies in human^{32, 33} and mice, suggest that the MAPK pathways need to be tightly regulated during development and that dysregulation can cause developmental disorders.

Heterozygous missense variants in TRAF7 have been ascertained as somatically acquired driver mutations in neoplastic tissue. Exome sequencing performed on intraneural perineuromas uncovered TRAF7 variants in the WD40 domain.³⁴ Clark et al. sequenced 300 meningiomas and found either missense or in-frame indel mutations in TRAF7 in 25% of samples, predominantly in the WD40 domain.³⁵ Goode et al. identified recurrent TRAF7 variants in adenomatoid tumors of the genital tract.³⁶ Thus, the nature and location of TRAF7 variants detected in tumor tissue mirror the TRAF7 germline events seen in our subjects. Intriguingly, the recurrent TRAF7 variants in somatic cancer, such as c.1562A>G (p.His521Arg) and c.1683C>A (p.Ser561Arg), have not been seen in our cohort or in the gnomAD database, possibly because these variants may be lethal if present as heterozygous constitutional variants. Notably, subject 3, the eldest patient in our cohort, has been diagnosed with a meningioma. Given that somatic changes in TRAF7 have been reported in patients with meningioma and mesothelioma,⁸ additional long-term follow-up of the younger subjects will be necessary to definitively determine whether germline TRAF7 variants predispose to meningioma or other tumor formation.

In conclusion, we propose that germline missense variants in TRAF7 are associated with a human phenotype characterized by developmental delay, heart defects, and dysmorphic features. The shared phenotypes among the patients, statistically significant enrichment of de novo TRAF7 variants in our exome cohort, high phylogenetic conservation of the altered amino acids, localization of variants to key functional domains, and reduced ERK1/2 activation resulting from the TRAF7 variants support the putative pathogenicity of these variants.

Declaration of Interests

The Department of Molecular and Human Genetics at Baylor College of Medicine derives revenue from molecular genetic testing offered at Baylor Genetics.

Acknowledgments

Notes

Published: June 28, 2018

Footnotes

References